21/09/2018 14:16, Gaëtan Rivet: > On Thu, Sep 20, 2018 at 06:17:13PM +0200, Thomas Monjalon wrote: > > 19/09/2018 18:03, Gaetan Rivet: > > > The eth device class can now parse a field name, > > > matching the eth_dev name with one passed as > > > > > > "class=eth,name=xxxxxx" > > > > I am not sure what is the purpose of the "name" property. > > I think we should not need it to choose a port by its ethdev name. > > rte_eth_dev_get_port_by_name seems pretty close.
Exact, this function already exists to get a port id by name. I think we should discourage the use of ethdev "internal" name. The point of the devargs is to match device with explicit known properties. > Which fields do you think should be proposed first as a getter on the eth > class? > Or do you have a list of fields the eth class should be restricted to? In other words, which ethdev properties can help to match a port? I think about "mac=" (which is already implemented in OVS devargs), and "representor=" (which requires a new field in ethdev). About representors, we could also match all representor ports of a switch (see rte_eth_switch_info). We could also have a property for the kernel netdev we are (or were) bound. Does it make sense? > > If you are thinking about a vdev, we can use the rte_device name (at bus > > level). > > This patch is only about eth class, it has no impact on buses. So we agree :)
